In federated learning, textual prompt tuning adapts Vision-Language Models (e.g., CLIP) by tuning lightweight input tokens (or prompts) on local client data, while keeping network weights frozen. After training, only the prompts are shared by the clients with the central server for aggregation. However, textual prompt tuning suffers from overfitting to known concepts, limiting its generalizability to unseen concepts. To address this limitation, we propose Multimodal Visual Prompt Tuning (FedMVP) that conditions the prompts on multimodal contextual information - derived from the input image and textual attribute features of a class. At the core of FedMVP is a PromptFormer module that synergistically aligns textual and visual features through a cross-attention mechanism. The dynamically generated multimodal visual prompts are then input to the frozen vision encoder of CLIP, and trained with a combination of CLIP similarity loss and a consistency loss. Extensive evaluation on 20 datasets, spanning three generalization settings, demonstrates that FedMVP not only preserves performance on in-distribution classes and domains, but also displays higher generalizability to unseen classes and domains, surpassing state-of-the-art methods by a notable margin of +1.57% - 2.26%. Code is available at https://github.com/mainaksingha01/FedMVP.

We delve into Open Domain Generalization (ODG), marked by domain and category shifts between training's labeled source and testing's unlabeled target domains. Existing solutions to ODG face limitations due to constrained generalizations of traditional CNN backbones and errors in detecting target open samples in the absence of prior knowledge. Addressing these pitfalls, we introduce ODG-CLIP, harnessing the semantic prowess of the vision-language model, CLIP. Our framework brings forth three primary innovations: Firstly, distinct from prevailing paradigms, we conceptualize ODG as a multi-class classification challenge encompassing both known and novel categories. Central to our approach is modeling a unique prompt tailored for detecting unknown class samples, and to train this, we employ a readily accessible stable diffusion model, elegantly generating proxy images for the open class. Secondly, aiming for domain-tailored classification (prompt) weights while ensuring a balance of precision and simplicity, we devise a novel visual stylecentric prompt learning mechanism. Finally, we infuse images with class-discriminative knowledge derived from the prompt space to augment the fidelity of CLIP's visual embeddings. We introduce a novel objective to safeguard the continuity of this infused semantic intel across domains, especially for the shared classes. Through rigorous testing on diverse datasets, covering closed and open-set DG contexts, ODG-CLIP demonstrates clear supremacy, consistently outpacing peers with performance boosts between 8%-16%. Code will be available at https://github.com/mainaksingha01/ODG-CLIP.

CRISPR/Cas9 (Clustered Regularly Interspaced Short Palindromic Repeats and CRISPR-associated protein 9) is a popular and effective two-component technology used for targeted genetic manipulation. It is currently the most versatile and accurate method of gene and genome editing, which benefits from a large variety of practical applications. For example, in biomedicine, it has been used in research related to cancer, virus infections, pathogen detection, and genetic diseases. Current CRISPR/Cas9 research is based on data-driven models for on- and off-target prediction as a cleavage may occur at non-target sequence locations. Nowadays, conventional machine learning and deep learning methods are applied on a regular basis to accurately predict on-target knockout efficacy and off-target profile of given single-guide RNAs (sgRNAs). In this paper, we present an overview and a comparative analysis of traditional machine learning and deep learning models used in CRISPR/Cas9. We highlight the key research challenges and directions associated with target activity prediction. We discuss recent advances in the sgRNA-DNA sequence encoding used in state-of-the-art on- and off-target prediction models. Furthermore, we present the most popular deep learning neural network architectures used in CRISPR/Cas9 prediction models. Finally, we summarize the existing challenges and discuss possible future investigations in the field of on- and off-target prediction. Our paper provides valuable support for academic and industrial researchers interested in the application of machine learning methods in the field of CRISPR/Cas9 genome editing.

Generalized zero-shot learning (GZSL) aims to train a model for classifying data samples under the condition that some output classes are unknown during supervised learning. To address this challenging task, GZSL leverages semantic information of the seen (source) and unseen (target) classes to bridge the gap between both seen and unseen classes. Since its introduction, many GZSL models have been formulated. In this review paper, we present a comprehensive review on GZSL. First, we provide an overview of GZSL including the problems and challenges. Then, we introduce a hierarchical categorization for the GZSL methods and discuss the representative methods in each category. In addition, we discuss the available benchmark data sets and applications of GZSL, along with a discussion on the research gaps and directions for future investigations.

In this article, a new feature selection (FS) algorithm, called simple, fast, and efficient (SFE), is proposed for high-dimensional datasets. The SFE algorithm performs its search process using a search agent and two operators: 1) nonselection and 2) selection. It comprises two phases: 1) exploration and 2) exploitation. In the exploration phase, the nonselection operator performs a global search in the entire problem search space for the irrelevant, redundant, trivial, and noisy features and changes the status of the features from selected mode to nonselected mode. In the exploitation phase, the selection operator searches the problem search space for the features with a high impact on the classification results and changes the status of the features from nonselected mode to selected mode. The proposed SFE is successful in FS from high-dimensional datasets. However, after reducing the dimensionality of a dataset, its performance cannot be increased significantly. In these situations, an evolutionary computational method could be used to find a more efficient subset of features in the new and reduced search space. To overcome this issue, this article proposes a hybrid algorithm, SFE-PSO (particle swarm optimization) to find an optimal feature subset. The efficiency and effectiveness of the SFE and the SFE-PSO for FS are compared on 40 high-dimensional datasets. Their performances were compared with six recently proposed FS algorithms. The results obtained indicate that the two proposed algorithms significantly outperform the other algorithms and can be used as efficient and effective algorithms in selecting features from high-dimensional datasets.

The automatic and accurate analysis of medical images (e.g., segmentation,detection, classification) are prerequisites for modern disease diagnosis and prognosis. Computer-aided diagnosis (CAD) systems empower accurate and effective detection of various diseases and timely treatment decisions. The past decade witnessed a spur in deep learning (DL)-based CADs showing outstanding performance across many health care applications. Medical imaging is hindered by multiple sources of uncertainty ranging fromnteasurement (aleatoric) errors, physiological variability, and limited medical knowledge (epistemic errors). However, uncertainty quantification (UQ) in most existing DL methods is insufficiently investigated, particularly in medical image analysis. Therefore, to address this gap, in this article, we propose a simple yet novel hierarchical attentive multilevel feature fusion model with an uncertainty-aware module for medical image classification coinedHercules. This approach is tested on several real medical image classification challenges. The proposedHerculesmodel consists of two main feature fusion blocks, where the former concentrates on attention-based fusion with uncertainty quantification module and the latter uses the raw features.Herculeswas evaluated across three medical imaging datasets, i.e., retinal OCT, lung CT, and chest X-ray.Herculesproduced the best classification accuracy in retinal OCT (94.21%), lung CT (99.59%), and chest X-ray (96.50%) datasets, respectively, against other state-of-the-art medical image classification methods.

Modern radiotherapy delivers treatment plans optimised on an individual patient level, using CT-based 3D models of patient anatomy. This optimisation is fundamentally based on simple assumptions about the relationship between radiation dose delivered to the cancer (increased dose will increase cancer control) and normal tissue (increased dose will increase rate of side effects). The details of these relationships are still not well understood, especially for radiation-induced toxicity. We propose a convolutional neural network based on multiple instance learning to analyse toxicity relationships for patients receiving pelvic radiotherapy. A dataset comprising of 315 patients were included in this study; with 3D dose distributions, pre-treatment CT scans with annotated abdominal structures, and patient-reported toxicity scores provided for each participant. In addition, we propose a novel mechanism for segregating the attentions over space and dose/imaging features independently for a better understanding of the anatomical distribution of toxicity. Quantitative and qualitative experiments were performed to evaluate the network performance. The proposed network could predict toxicity with 80% accuracy. Attention analysis over space demonstrated that there was a significant association between radiation dose to the anterior and right iliac of the abdomen and patient-reported toxicity. Experimental results showed that the proposed network had outstanding performance for toxicity prediction, localisation and explanation with the ability of generalisation for an unseen dataset.

Nonlinear model predictive control has been used in motion cueing algorithms recently to consider the nonlinear dynamics model of the system. The entire motion cueing algorithm indexes, including the physical and dynamical constraints of the actuators and physical constraints of passive joints, can be controlled with precision using nonlinear model predictive control. However, several weighting parameters in the nonlinear model predictive control-based motion cueing algorithm (including driving sensation, motion description of the actuators, and passive joints) require proper and laborious tuning to attain an optimal design structure. In this work, the optimal weighting parameters of a nonlinear predictive control-based motion cueing algorithm model are calculated using cascade optimisation and human interaction. A cascade optimisation method consisting of a particle swarm optimisation and genetic algorithm is designed to identify the best weighting parameters compared to those from one optimiser. In addition, the human decision-making units are added to the two-level cascade optimiser to determine the best solution from a Pareto front. The proposed cascade optimiser decreases the run-time with better extraction of the optimal weighting parameters to increase the motion fidelity compared to a single optimiser. It should be noted that the proposed methodology is applied along longitudinal channel. While the same methodology can be applied along lateral, heave and yaw channels for further evaluation of the proposed method. The proposed model is simulated utilising the MATLAB software and the results prove the efficiency of the newly proposed model compared to those from the previous single optimiser in reproducing more accurate motion signals with better usage of the driving motion platform workspace.

Abstract Coronary artery disease (CAD) is the leading cause of morbidity and death worldwide. Invasive coronary angiography is the most accurate technique for diagnosing CAD, but is invasive and costly. Hence, analytical methods such as machine learning and data mining techniques are becoming increasingly more popular. Although physicians need to know which arteries are stenotic, most of the researchers focus only on CAD detection and few studies have investigated stenosis of the right coronary artery (RCA), left circumflex (LCX) artery and left anterior descending (LAD) artery separately. Meanwhile, most of the datasets in this field are noisy (data uncertainty). However, to the best of our knowledge, there is no study conducted to address this important problem. This study uses the extension of the Z‐Alizadeh Sani dataset, containing 303 records with 54 features. A new feature selection algorithm is proposed in this work. Meanwhile, by discretization of data, we also handle the uncertainty in CAD prediction. To the best of our knowledge, this is the first study attempted to handle uncertainty in CAD prediction. Finally, the genetic algorithm (GA) is used to determine the hyper‐parameters of the support vector machine (SVM) kernels. We have achieved high accuracy for the stenosis diagnosis of each main coronary artery. The results of this study can aid the clinicians to validate their manual stenosis diagnosis of RCA, LCX and LAD coronary arteries.

Abstract Coronary artery disease (CAD) is one of the major causes of mortality worldwide. Knowledge about risk factors that increase the probability of developing CAD can help to understand the disease better and assist in its treatment. Recently, modern computer‐aided approaches have been used for the prediction and diagnosis of diseases. Swarm intelligence algorithms like particle swarm optimization (PSO) have demonstrated great performance in solving different optimization problems. As rule discovery can be modelled as an optimization problem, it can be mapped to an optimization problem and solved by means of an evolutionary algorithm like PSO. An approach for discovering classification rules of CAD is proposed. The work is based on the real‐world CAD data set and aims at the detection of this disease by producing the accurate and effective rules. The proposed algorithm is a hybrid binary‐real PSO, which includes the combination of categorical and numerical encoding of a particle and a different approach for calculating the velocity of particles. The rules were developed from randomly generated particles, which take random values in the range of each attribute in the rule. Two different feature selection methods based on multi‐objective evolutionary search and PSO were applied on the data set, and the most relevant features were selected by the algorithms. The accuracy of two different rule sets were evaluated. The rule set with 11 features obtained more accurate results than the rule set with 13 features. Our results show that the proposed approach has the ability to produce effective rules with highest accuracy for the detection of CAD.

This paper proposes a new approach to produce classification rules based on evolutionary computation with novel crossover and mutation operators customized for execution on graphics processing unit (GPU). Also, a novel method is presented to define the fitness function, i.e. the function which measures quantitatively the accuracy of the rule. The proposed fitness function is benefited from parallelism due to the parallel execution of data instances. To this end, two novel concepts; coverage matrix and reduction vectors are used and an altered form of the reduction vector is compared with previous works. Our CUDA program performs operations on coverage matrix and reduction vector in parallel. Also these data structures are used for evaluation of fitness function and calculation of genetic operators in parallel. We proposed a vector called average coverage to handle crossover and mutation properly. Our proposed method obtained a maximum accuracy of 99.74% for Hepatitis C Virus (HCV) dataset, 95.73% for Poker dataset, and 100% for COVID-19 dataset. Our speedup is higher than 20% for HCV and COVID-19, and 50% for Poker, compared to using single core processors.

Credit scoring (CS) is an effective and crucial approach used for risk management in banks and other financial institutions. It provides appropriate guidance on granting loans and reduces risks in the financial area. Hence, companies and banks are trying to use novel automated solutions to deal with CS challenge to protect their own finances and customers. Nowadays, different machine learning (ML) and data mining (DM) algorithms have been used to improve various aspects of CS prediction. In this paper, we introduce a novel methodology, named Deep Genetic Hierarchical Network of Learners (DGHNL). The proposed methodology comprises different types of learners, including Support Vector Machines (SVM), k-Nearest Neighbors (kNN), Probabilistic Neural Networks (PNN), and fuzzy systems. The Statlog German (1000 instances) credit approval dataset available in the UCI machine learning repository is used to test the effectiveness of our model in the CS domain. Our DGHNL model encompasses five kinds of learners, two kinds of data normalization procedures, two extraction of features methods, three kinds of kernel functions, and three kinds of parameter optimizations. Furthermore, the model applies deep learning, ensemble learning, supervised training, layered learning, genetic selection of features (attributes), genetic optimization of learners parameters, and novel genetic layered training (selection of learners) approaches used along with the cross-validation (CV) training-testing method (stratified 10-fold). The novelty of our approach relies on a proper flow and fusion of information (DGHNL structure and its optimization). We show that the proposed DGHNL model with a 29-layer structure is capable to achieve the prediction accuracy of 94.60% (54 errors per 1000 classifications) for the Statlog German credit approval data. It is the best prediction performance for this well-known credit scoring dataset, compared to the existing work in the field.

Abstract Image classification (categorization) can be considered as one of the most breathtaking domains of contemporary research. Indeed, people cannot hide their faces and related lineaments since it is highly needed for daily communications. Therefore, face recognition is extensively used in biometric applications for security and personnel attendance control. In this study, a novel face recognition method based on perceptual hash is presented. The proposed perceptual hash is utilized for preprocessing and feature extraction phases. Discrete Wavelet Transform (DWT) and a novel graph based binary pattern, called quintet triple binary pattern (QTBP), are used. Meanwhile, the K-Nearest Neighbors (KNN) and Support Vector Machine (SVM) algorithms are employed for classification task. The proposed face recognition method is tested on five well-known face datasets: AT&T, Face94, CIE, AR and LFW. Our proposed method achieved 100.0% classification accuracy for the AT&T, Face94 and CIE datasets, 99.4% for AR dataset and 97.1% classification accuracy for the LFW dataset. The time cost of the proposed method isO(nlogn). The obtained results and comparisons distinctly indicate that our proposed has a very good classification capability with short execution time.

In an opinionated long review, there may be several targets described by different potential terms. Traditional review-level techniques for Persian sentiment analysis addressed the problem using a one-method-fits-all solution in which the overall polarity of a review is calculated using all its opinionated words without considering their target. In this article, a new method is proposed, which first decomposes a long review into its constituent sentences and then detects the main target of each sentence. In the next step, five policies, including most occurring first (MOF), most general first (MGF), most specific first (MSF), first occurring first (FOF), and last occurring first (LOF), are proposed to come up with the main target of the review. Finally, using the part-of-speech (POS) tags, potential terms in the sentences are specified and a comprehensive sentiment lexicon is employed to compute the polarity of the sentences. In order to evaluate the proposed method, three data sets of user reviews about different topics, including digital equipment, hotels, and movies, are created as no previous study addressed the problem of target identification in the Persian language. The results of comparing the proposed method with a state-of-the-art lexicon-based method show that specifying the main targets of reviews can improve the performance of the systems about 17% and 12% in terms of accuracy and F1-measure. Moreover, the proposed method using the MGF policy achieves the best performance in finding the main target of reviews, while for finding the ultimate polarity of reviews, the MOF outperforms other policies.